On the Automation of Code Review Tasks Through Cross-Task Knowledge Distillation

Code review is essential for ensuring code quality and minimizing errors, but it is often complex, subjective, and time-consuming. Existing research treated code review tasks (i.e., quality estimation, comment generation, and code refinement) separately, despite their interconnections. We propose a novel deep learning architecture using cross-task knowledge distillation to address these tasks simultaneously. Our approach outperforms state-of-the-art methods in comment generation and code refinement, achieving superior BLEU and CodeBLEU scores.